Why Your Non-Studio Recording Sounds Off

You do not need a $400 microphone to get listenable voice audio. You need a quiet-ish room, decent mic technique, and a cleanup chain that removes noise without turning your voice into a underwater robot. Most bad recordings fail for one of four reasons: steady hiss or hum, room echo, inconsistent distance from the mic, or digital clipping. Noise removal can help with the first one. It cannot rebuild lost detail or remove a hollow room sound completely. So think of cleanup as damage control plus polish.

This guide is for a podcast, YouTube voiceover, Zoom call, gaming clip, or phone voice memo that already exists. You will learn a repeatable order of operations, concrete starting settings, and the tradeoffs that matter.

Step 0: Save the Original and Find Room Tone

Duplicate the file before you do anything. Keep one untouched master. Then find 5-10 seconds where nobody is talking. That is your room tone. If the recording has no silence, find the quietest pause between words. You will use this sample to teach noise reduction what the noise sounds like.

Listen with headphones, not laptop speakers. You need to hear the difference between constant noise and intermittent noise. Constant noise is easier to remove. Keyboard clicks, coughs, and chair creaks need manual editing or a gate.

Step 1: Remove Rumble and Hum First

High-pass filter

Apply a high-pass filter at 80 Hz for a deep male voice, 100 Hz for a female or higher voice, and 120 Hz for a thin or child voice. Use a slope of 12 or 24 dB per octave. This removes desk thumps, AC rumble, and mic handling noise without hurting intelligibility. If the voice sounds thin after, lower the frequency by 10-20 Hz.

Notch out electrical hum

If you hear a steady low buzz, it is probably 60 Hz in the US or 50 Hz in Europe. Use a narrow notch filter, Q of 20-30, and cut 6-18 dB at the fundamental. Then check the harmonics: 120 Hz and 180 Hz for 60 Hz hum, or 100 Hz and 200 Hz for 50 Hz hum. Do not cut more than you need.

Manual edit the obvious stuff

Before any noise reduction, delete or mute clicks, lip smacks, and loud breaths. Do not delete the space entirely; use a short fade so the audio does not sound choppy.

Step 2: Noise Reduction Without the Watery Artifact

In Audacity, select the room tone, open Noise Reduction, and click Get Noise Profile. Then select the whole track and open the effect again. Start with 6-12 dB of reduction, Sensitivity 6.00, and Frequency Smoothing 3. Preview a loud and a quiet section. If you hear metallic chirps or a watery tail, undo and try 4-6 dB. Two gentle passes often sound better than one aggressive pass.

If you do not want to learn a full editor, a free online option like ClearVoiceHQ (clearvoicehq.com) can handle the noise-removal pass for you. Upload the file, let it process, then download and continue with the EQ and compression steps below. Always compare the cleaned file to the original on headphones.

Should you use a gate?

A noise gate can silence background noise between phrases, but it is easy to misuse. Set the threshold just above your noise floor, usually -45 to -35 dB in a quiet room or -30 dB in a noisy one. Use an attack of 5-10 ms, hold of 50-100 ms, and release of 150-300 ms. Set the range to -12 or -18 dB instead of -infinity so the silence still has a little room tone. If the gate makes speech cut in and out, lower the threshold or turn it off.

Step 3: EQ for Speech, Not Music

After noise reduction, use EQ to shape the voice. A good starting point for spoken word:

Step 4: Compress and Set Loudness

Compression makes quiet words easier to hear. Start with a ratio of 2:1 to 4:1, threshold around -18 to -12 dB, attack 10-20 ms, and release 80-150 ms. Add makeup gain so the loudest peaks hit around -6 to -3 dB. Avoid heavy limiting at this stage.

For final loudness, aim for -16 LUFS mono for podcasts and -14 LUFS for YouTube stereo. Set true peak at -1 dBTP. For social media voice memos, -14 to -12 LUFS is fine.

Step 5: When AI Noise Removal Is the Right Tool

AI noise removal, including ClearVoice, is best for steady noise: fans, air conditioners, street rumble, hiss from a cheap preamp, or a laptop fan in a Zoom call. It is also useful when you have many clips and no time to dial in a noise profile for each one. The tradeoff is that aggressive AI can soften consonants, add a slight underwater texture, or remove quiet breaths that make speech sound natural. Use it as one pass, not five.

If the recording has heavy reverb, AI may reduce it but not eliminate it. If the recording is clipped, AI cannot restore the missing peaks. In those cases, re-record or lean on manual editing, EQ, and compression.

A Copy-and-Paste Cleanup Chain

  1. Duplicate the original file. Keep a backup.
  2. Find room tone and run noise reduction at 6-12 dB. Check for artifacts.
  3. High-pass filter at 80-120 Hz.
  4. Notch out 50 or 60 Hz hum and its harmonics if needed.
  5. Manually edit clicks, smacks, and loud breaths.
  6. Apply EQ cuts and presence boost.
  7. Compress at 3:1 with -18 dB threshold, then de-ess.
  8. Normalize to -16 LUFS mono or -14 LUFS stereo, true peak -1 dBTP.
  9. Export as WAV for editing or 192-320 kbps MP3/AAC for upload.

How to Get a Cleaner Recording Next Time

The best cleanup is a better capture. Move the mic 6-12 inches from your mouth, slightly off-axis. Speak across the mic, not directly into it. Put a blanket, duvet, or moving blanket behind you and in front of hard walls. Turn off fans, HVAC, and notifications. Record 10 seconds of room tone before every take. Set gain so loud peaks hit -12 to -6 dB, never 0 dB. If you use a phone, record in a quiet closet and keep the phone on a stand, not in your hand.

Do those things and your cleanup chain becomes a 5-minute polish instead of a rescue mission. You still will not sound like a studio, but you will sound clear, consistent, and easy to listen to, which is what your audience actually cares about.